Data Centers

Optical processor startup claims to outperform GPUs for AI workloads

As performance increases in traditional processors have stagnated in recent years, development of alternative processors is accelerating.

Building a slide deck, pitch, or presentation? Here are the big takeaways:
  • Fathom Computing is building an optical computer for neural networks, intended for artificial intelligence use cases.
  • Optical processors are receiving renewed interest as performance increases of transistor-based CPUs have stagnated.

Fathom Computing, a startup founded by brothers William and Michael Andregg, is aiming to design an optical computer for neural networks. As performance increases in traditional transistor-based CPUs have relatively stagnated over the last few years, alternative computing paradigms such as quantum computers have been gaining traction. While optical computers are not a new concept—Coherent Optical Computers was published in 1972—they have been relegated to university research laboratories for decades.

The design of the Fathom prototype performs mathematical operations by encoding numbers into light, according to this profile in Wired. The light is then passed through a series of lenses and other optical components. The measured result of this process is the calculated result of the operation.

The Fathom prototype is not a general-purpose processor, it is designed to compute specific types of linear algebra operations. Specifically, Fathom is targeting the long short-term memory type of recurrent neural networks, as well as the non-recurrent feedforward neural network. This mirrors trends in quantum computers, as systems produced by D-Wave are similarly targeted toward quantum annealing, rather than general processing.

SEE: IT leader's guide to the future of quantum computing (Tech Pro Research)

In contrast to quantum computers, optical computers would theoretically require less power to operate, as the devices principally run on photonic (light) signals, rather than electrical signals. Accordingly, electrical resistance generates waste heat, requiring the use of aggressive cooling systems for large data centers. Quantum computers use liquid helium cooling systems, which will continue to be viable for some time following the 2016 discovery of a 1.53 km³ reserve in Tanzania. Though specific temperatures were not discussed, the Fathom prototype is noted in the Wired profile as encountering difficulty when too cold.

In a recent blog post, Fathom indicated that they are still two years away from launching, but that they expect their platform to "significantly outperform state-of-the-art GPUs." The first systems will be available as a cloud service for researchers working in artificial intelligence.

Due to the specialized nature of both optical and quantum computers, they are more likely to be used in conjunction with traditional computers than to serve as an outright replacement. The relationship between a standard x86-64, ARM, or POWER processor and these new paradigms of computing more closely resembles the difference between CPU and GPU in current systems.

Fathom is registered as a public benefit corporation—the company's mission statement notes that their goal is to "ensure that the benefits of these increased capabilities serve to improve all lives." The company has received backing from Android co-founder Andy Rubin's Playground venture capital firm, with which it shares an office. Other organizations are entering the nascent field of optical computing, including Hewlett Packard Enterprise and French startup LightOn.

Also see

neural.jpg
Image: iStockphoto/sakkmesterke

About James Sanders

James Sanders is a Writer for TechRepublic. Since 2013, he has been a regular contributor to TechRepublic and Tech Pro Research.

Editor's Picks

Free Newsletters, In your Inbox